Recognition of Spoken Arabic Digits Using Neural Predictive Hidden Markov Models

Recognition of Spoken Arabic Digits Using Neural Predictive Hidden Markov Models

Rafik Djemili, Mouldi Bedda, and Hocine Bourouba

Automatic and Signals Laboratory of Annaba, Badji Mokhtar University, Algeria

 

Abstract: In this study, we propose an algorithm for Arabic isolated digit recognition.  The algorithm is  based   on extracting  acoustical  features  from  the  speech  signal  and  using  them  as  input  to  multi-layer  perceptrons  neural  networks.  Each word in the vocabulary digits (0 to 9) is associated with a network. The networks are implemented as predictors for the speech samples for certain duration of time. The back-propagation algorithm is used to train the networks. The hidden markov model (HMM) is implemented to extract temporal features (states) for the speech signal. The input vector to the networks consists of twelve mel frequency cepstral coefficients, log of the energy, and five elements representing the state. Our results show that we are able to reduce the word error rate comparing with an HMM word recognition system.

 

Keywords: Speech recognition, hidden Markov models, artificial neural networks, hybrid HMM/MLP.

 

Received September 15, 2003; accepted January 19, 2004

Full Text

  
Read 8377 times Last modified on Wednesday, 20 January 2010 03:52
Share
Top
We use cookies to improve our website. By continuing to use this website, you are giving consent to cookies being used. More details…